The following patent applications, herein incorporated by reference, describe the background of this invention: Provisional Patent Application Ser. No. 60/267,260, inventor Thomas N. Horsky, filed Feb. 7, 2001, entitled Ion Source for Ion Implantation; Provisional Patent Application Ser. No. 60/257,322, inventor Thomas N. Horsky, filed Dec. 19, 2000, entitled Ion Implantation; PCT Application Serial Number US00/33786, inventor Thomas N. Horsky, filed Dec. 13, 2000, entitled Ion Implantation Ion Source, System and Method and filed Nov. 30, 2000, having the same reference. The referenced patent, for U.S. Purposes, is a continuation in part of my U.S. Provisional Applications 60/170,473 filed Dec. 13, 1999 60/170,473, now expired.
Background: Ion Implantation
Ion implantation has been a key technology in semiconductor device manufacturing for more than twenty years, and is currently used to fabricate the p-n junctions in transistors, particularly for CMOS devices such as memory and logic chips. By creating positively-charged ions containing the dopant elements (for example, 75As, 11B, 115In, 31P, or 121Sb) required for fabricating the transistors in, for example, silicon substrates, the ion implanters can selectively control both the energy (hence implantation depth) and ion current (hence dose) introduced into the transistor structures. Ion implanters have traditionally used ion sources which generate ribbon beams of up to about 50 mm in length; these beams are transported to the substrate and the required dose and dose uniformity is accomplished by electromagnetic scanning of the ribbon across the substrate, mechanical scanning of the substrate across the beam, or both.
With the recent advent of 300 mm-diameter silicon substrates in chip manufacturing, there has been a keen interest in producing ribbons of larger extent than has heretofore been possible with conventional ion implanter designs, in order to increase wafer throughput when using these larger substrates. Taller ribbon beams enable higher dose rates, since more ion current can be transported through the implanter beam line due to reduced space charge blowup of the extended ribbon beam. Many of these new implanter designs also incorporate a serial (one wafer at a time) process chamber, which offers high tilt capability (e.g., up to 60 degrees from substrate normal). The ion beam is typically electromagnetically scanned across the wafer, which is mechanically scanned in the orthogonal direction, to ensure dose uniformity. In order to meet implant dose uniformity and repeatability specifications, the ion beam must have excellent angular and spatial uniformity (angular uniformity of beam on wafer of <1 deg, for example). The production of beams possessing these characteristics imposes severe constraints on the beam transport optics of the implanter, and the use of large-emittance plasma-based ion sources often results in increased beam diameter and beam angular divergence, causing beam loss during transport due to vignetting of the beam by apertures within the beam line. Currently, the generation of high current ion beams at low (<2 keV) energy is problematic in serial implanters, such that wafer throughput is unacceptably low for certain low-energy implants (for example, in the creation of source and drain structures in leading-edge CMOS processes). Similar transport problems exist for batch implanters (processing many wafers mounted on a spinning disk), particularly at low beam energies.
While it is possible to design beam transport optics which are nearly aberration-free, the beam characteristics (spatial extent and angular divergence) are nonetheless determined to a large extent by the emittance properties of the ion source (i.e., the beam properties at ion extraction which determine the extent to which the implanter optics can focus and control the beam as emitted from the ion source). Arc-discharge plasma sources which are currently in use have poor emittance, and therefore limit the ability of ion implanters to produce well-focused, collimated, and controllable ion beams.
Background: Ion Implantation Sources
The standard ion source technology of the implanter industry is the Enhanced Bernas source. As illustrated in FIG. 1, this is an arc discharge source which incorporates a reflex geometry: a hot filament cathode immersed in the ionization chamber (where the dopant feed gas resides) emits thermionic electrons confined by a magnetic field, and are reflected from an anticathode located at the opposite end of the chamber. Thus, the electrons execute helical trajectories between the cathode and anticathode, and generate a high-density plasma (on the order of 1012 ions/cm2). This so-called “plasma column” is parallel to an ion extraction aperture slot from which the ions are extracted by beam-forming optics. By generating a high-density plasma and sustaining discharge currents as high as 10A, the Enhanced Bernas source efficiently dissociates tightly-bound molecular species such as BF3. However, the emittance of this source is large due to the following plasma-related effects:                1) The plasma potential (typically about 5 V) introduces a random component of velocity to the ions, which directly translates into increased angular dispersion of the extracted ions.        2) The temperature of the ions and electrons within the plasma can reach 10,000 K, introducing a thermal velocity which adds to (1), and also introduces an energy spread of several eV to the ions (according to a Maxwell-Boltzmann distribution), making the beam exhibit chromatic aberrations.        3) Coulomb scattering between the ions in the plasma introduces an additional non-thermal spread in the ion energy.        4) A high extracted current density is needed due to a predominance of unwanted ions (i.e., fragments such as BF+, BF2+, and F− in a BF3 plasma), increasing space-charge forces at extraction and causing emittance growth.        5) The presence of a strong magnetic field, required for operation of all arc discharge sources, causes beam deflection and hence further emittance growth of the extracted ion beam, especially at low beam energy.        6) High-frequency noise present in the plasma is propagated into the beam as high-frequency fluctuations in beam current and in beam potential. This time-varying beam potential makes charge compensation in the beam plasma difficult to maintain, since it can cause a significant steady or even abrupt loss of the low-energy electrons which normally orbit the beam (being trapped by the positive beam potential), leading to space-charge blowup of the ion beam.        7) The ion extraction aperture cannot be significantly elongated beyond, say, 75 mm (typical length is between 20 mm and 50 mm), since this requires a significant elongation of the plasma column. Bernas sources become unstable if the separation between cathode and anticathode is large, and larger cathode-anticathode separations requires a much higher arc discharge current in order to maintain a stable plasma, increasing power consumption.Background: Ion Deceleration        
Ion implanters or conventional design exhibit poor transmission of low-energy boron at energies below a few keV, with the result that these boron beam currents are too small to be cost-effective in manufacturing semiconductor chips using sub-0.18 micron design rules. Next-generation implanters which have been long-in-planning, and which were introduced into the capital equipment market within the last few years incorporate a different principle of ion optics, attempting to solve this low-energy transmission problem. To counter the effects of space charge repulsion between ions, which dominates beam transport at low energies, a so-called “decel” (i.e. deceleration) approach has been developed to allow the ion beam to be extracted and transported through the implanter at a higher energy than the desired implantation energy so that space charge effects are not so detrimental, and by introducing a deceleration stage late in the beam-line, but upstream from the wafer target, reducing the ions to the desired implant energy as the ions approach the wafer target. For example, the ion beam can be extracted and transported at 2 keV, but decelerated to 500 eV before the ions reach the wafer, achieving a much higher beam current than is obtainable with space-charge-limited beams in beam lines of a conventional, non-deceleration design. Unfortunately, this method of employing deceleration still has posed significant problems which have detracted from its production-worthiness. As the ion beam passes through the deceleration lens to the wafer, the ion beam becomes spatially non-uniform to a great degree, and the ions impact the substrate with a wide distribution of angles of incidence relative to the wafer surface, with potential so-called channeling effects. The spatial and angular dose uniformity of a decelerated beam is typically much worse than in conventional, non-deceleration ion implantation. This makes it difficult to achieve a uniform dose, and introduces the need to take other steps which affects cost and throughput. Compounding the problem is the fact that the grossly non-uniform profile of the ion beam also interferes significantly with accurate dosimetry of the implant, since ion implanters typically sample only a portion of an ion beam at or behind the plane of the wafer. Dosimetry is used to control the degree of implant within a desired range. The accuracy problems with dosimetry produced by partial sampling of a severely extended and non-uniform distribution of ion current in the beam of an acceleration/deceleration implantation system thus also affects the accuracy of the implant, the capital cost of the implant system, the quality of the wafers, and throughput of the system.
Another, quite different approach for shallow, low energy implants has been proposed (but not implemented in current production) it is that of using molecular ion beams (having clusters of the dopant atom of interest) in conventional implanters that do not have a deceleration stage. Decaborane is one example of such molecular material.
Chip manufactures are currently moving to 300 mm-diameter silicon substrates for fabricating Complimentary Metal-Oxide-Semiconductor (CMOS) memory and logic chips to reduce manufacturing cost over that attainable with 200 mm substrates. Though such a shift in wafer size requires building new factories populated with new semiconductor manufacturing tools for processing the larger-diameter wafers, the potential reduction per die is about a factor of two. Thus, the expenditure of billions of US dollars for these facilities has been hoped to enable lower-cost manufacturing, and ultimately a huge competitive advantage for volume manufacturing of both commodity and leading-edge semiconductor chips. Such a cost reduction can only be fully realized if the throughput of wafer units of the fab tools (the tools the fabrication facility) is the same for 300 mm as 200 mm substrates, which been to some extent been assumed would be the case. Unfortunately, in the case ion implantation to fabricate ultra-shallow (and ultra high density) semiconductor junctions, even the last acceleration/deceleration implanters continue to be dose-rate-limited in their wafer throughput, so that there has been essentially little or no net increase in productivity of semiconductor dies by use of the lager wafers. This is a potentially difficult situation for the chip manufacturer: if many more implanters must be put into production to make up their reduced output, the potential cost reduction per die sought by use of the larger wafer geometry cannot be realized due to the increased cost of performing these critical implants (more investment in capital equipment, fab floor space, maintenance cost, etc.).
Background: Ion Doping
Over the last decade, implantation systems have been developed for the ion implantation of very large substrates from which flat-panel displays are manufactured. These “Ion Doping” systems deliver long ribbon ion beams to the glass or quartz substrates, which are typically mechanically scanned across a stationary ion beam. The substrate dimension can be as large as a meter, and so the ion ribbon beam must likewise be long enough to ensure uniform doping (typically wider than the substrate). In order to generate such long ribbon beams, large-volume “bucket” sources are used. Bucket sources in a rectangular or cylindrical geometry are chambers surrounded by ail array of permanent magnets which provide magnetic confinement for the enclosed plasma through the creation of cusp magnetic fields. The plasma is generated by one or more RF antennas which couple RF power to the plasma. An extraction lens forms the ribbon beam from the large-diameter source.
Because of the size of the ion doping system, mass analysis is not used, therefore all ion species created in the bucket source are transported to and implanted into the substrate. This creates many process-related problems including variations in ion implantation depth, and also the implantation of unwanted species. Bucket sources are also particularly susceptible to the accumulation of deposits within their large ionization volume, hence the potential of severe cross-contamination between n- and p-type dopants requires the use of dedicated-use ion doping systems: the user must purchase one tool for p-type (e.g., boron from diborane gas) and a second complete tool for n-type (e.g., phosphorus from phosphene gas) dopants. This requirement not only doubles the customer's capital equipment costs, but substantially increases the risk of reduced product yield, since moving the substrates between systems requires further wafer handling steps and increased exposure of the substrates to atmosphere.
Thus, the prior art bucket source technology suffers from the following limitations:                (1) Large footprint (width, height and length).        (2) High degree of expense and complexity.        (3) Low ion production efficiency due to the loss of B (from B2H6 feed gas) and P (from PH3 feed gas) to the walls of the ion source due to the very large wall surface area and large volume of the source.        (4) Contamination and particulate problems associated with the rapid accumulation of deposits within the ion source associated with (3), reducing product yield.        (5) Production of many unwanted ions which are implanted into the substrate, resulting in a lack of implantation process control and a concomitant degradation of device characteristics. For example, significant fractions of H30  and BHx+, as well as B2Hx+, are produced in the B2H6 plasma commonly used to implant boron, a p-type dopant.        (6) Implantation of large currents of H30  (a result of (5)) during the implantation process limits attainable dose rate and hence throughput, since the total ion current delivered to the substrate must be held below a certain limit to prevent overheating of the substrate.        